Summary

dbt computes a resolved dependency graph into target/manifest.json, but it was never ingested, so a project's model and source lineage stayed invisible to the index even though dbt had already calculated it. This adds an opt-in MCP tool ingest_dbt_manifest {project, manifest_path}:

• model / seed / snapshot become Model / Seed / Snapshot nodes; source becomes a Source node, each keyed by dbt unique_id.
• Each node's depends_on.nodes becomes a DEPENDS_ON edge (model-to-model and model-to-source).
• test and analysis resources are skipped, so the graph keeps only lineage-bearing nodes.

It runs after index_repository against an existing store, so it augments rather than fabricates a graph, and is independent of index mode.

No schema change: reuses cbm_store_upsert_node / cbm_store_insert_edge and the existing DEPENDS_ON edge type.

Tests: tool_ingest_dbt_manifest_issue576 in tests/test_mcp.c, asserting 5 models + 2 sources + 5 DEPENDS_ON edges from a fixture manifest, with the test node skipped.

Fixes #576.

Related PRs

This is one of three PRs that split the SQL + dbt graph-indexing work to keep each under the one-issue-per-PR contributing rule. They share the same extraction and registry plumbing, so they are one logical change reviewed as a set:

• #582 : SQL DDL, first-class Table / View nodes + FROM/JOIN lineage (#574).
• #584 : dbt Jinja from raw .sql, Model / Macro nodes + ref() / source() lineage (#575).
• #583 (this PR): dbt manifest ingestion, Model / Source nodes + DEPENDS_ON lineage (#576).